How Ethereum Upgrades Affected L2s

The original idea was simple:
L2s were supposed to be "branded shards" of Ethereum — external execution layers inheriting L1 security.

But now L1 is already scaling on its own:

•  
  Fees are low 
• 
  GasLimit continues to grow 
• 
  Many L2s are unable or unwilling to become fully trustless.

This is a very important point.
In practice, many rollups are separate networks with a bridge to Ethereum, not an extension of L1.

A separate issue — liquidity fragmentation

Instead of a unified ecosystem, we ended up with many different ones that can't always hold their ground under current market conditions:
— Arbitrum
— Optimism
— Base
— zkSync
— Starknet

The more L2s there are, the greater the dependence on bridges — and the recent Kelp incident reinforces another key point: Bridges have become one of the largest sources of capital loss: total losses from bridge hacks amount to ~$1.83B.

An updated TVL snapshot across major L2 ecosystems shows a synchronized decline in sector liquidity:

Base: $5.38B → $4.31B (−19.9%)
Arbitrum: $4.07B → $1.51B (−62.9%)
Optimism: $415M → $342M (−17.6%)
Linea: $1.53B → $35M (−97.7%)
Starknet: $219M → $200M (−8.7%)
zkSync: $50M → $18M (−64.0%)

Overall, there is an uneven but sustained TVL decline across virtually all major L2 segments, with the depth of the drop varying significantly by ecosystem. The sharpest corrections are seen in Arbitrum, zkSync, and Linea, pointing to high liquidity sensitivity to incentive programs and short-term reward mechanics.

Taken together, the data supports a structural thesis: the L2 segment does not demonstrate sustainable capital accumulation between cycles, but is instead characterized by high liquidity rotation across networks and dependence on short-term incentives. This reinforces the broader picture of capital fragmentation in Ethereum's rollup ecosystem and undermines the "unified liquidity pool" effect that early L2 development models envisioned.